In datacenter fabrics, like an application centric infrastructure (ACI), an important troubleshooting step is to determine whether traffic forwarding issues, like drops, are caused by the datacenter fabric. Traditional packet counting techniques rely on comparing the egress and ingress counts for a given flow in order to determine whether the fabric is mis-forwarding the traffic. However, this is extremely difficult to achieve. Traffic that is constantly flowing is counted as part of the ingress and egress counts cannot be collected in an atomic fashion. In order to carry out granular atomic counting, interesting traffic must be identified ahead of time through changes in the configuration. The operator needs to know which flows need troubleshooting. Such a process renders the counting reactive in nature in that an operator needs to first detect a problem and then configure the required tenant parameters to atomically count packets.